基于变分模态分解的中国主要集装箱港口碳排放组合预测

    Carbon emission combined prediction of main ports in China based on variational modal decomposition

    • 摘要: 港口作为耗能和温室气体排放大户,研究其碳排放趋势对推进我国绿色生态港口建设至关重要。考虑到港口碳排放量波动具有多尺度特征,文章以中国主要集装箱港口为对象,构建了融合变分模态分解(VMD)-小波神经网络(WNN)-遗传算法(GA)-反向传播神经网络(BPNN)的多尺度组合预测模型。基于分解-分项预测-集成预测思想,采用VMD将碳排放量序列分解为多个模态分量;根据分量波动特征分为低、中、高频项和趋势项,分别优选预测方法实现分项预测;利用分项预测值完成集成预测并分析预测效果。实例应用表明,与现有预测模型相比,文章构建的多尺度组合预测模型能显著提高港口碳排放量预测精度,揭示港口碳排放量内在多尺度特征,有利于从能源技术、季节、突发事件等尺度制定针对性的碳减排策略。

       

      Abstract: As a major energy-consuming and greenhouse gas emitter,the study of carbon emission trends in ports is crucial for promoting the construction of green and ecological ports in China.Considering the multi-scale characteristics of port carbon emission fluctuations,this paper takes the major container ports in China as the objects,and constructs a multi-scale combined forecasting model integrating Variational Modal Decomposition-Wavelet Neural Network-Genetic Algorithm-Back Propagation Neural Network(VMD-WNN-GA-BPNN).Based on the idea of decomposition-subsequence forecastingensemble forecasting,the carbon emissions series are decomposed into multiple modal components by using VMD.The components are classified into low,medium and high frequency and trend terms according to their fluctuation characteristics,then,this paper optimizes forecast method respectively to achieve itemized forecast,and completes an integrated forecast and analyzes the effect of the forecast using the sub-predicted values.Example applications show that,compared with the existing prediction models,the multi-scale combined prediction model constructed in the paper can significantly improve the prediction accuracy of port carbon emissions and reveal the intrinsic multi-scale characteristics of port carbon emissions,which is conducive to the formulation of targeted carbon emission reduction strategies from the scales of energy technologies,seasons,and emergencies.

       

    /

    返回文章
    返回